The Application of Machine Learning Models to Predict Stillbirths
Abstract
1. Introduction
2. Material and Method
2.1. Study Design
2.2. Participant Selection and Data Collection
2.3. Machine Learning Models
- Accuracy: The overall proportion of correct predictions (both true positives and true negatives) made by the model.
- Area Under the Curve (AUC): The area under the receiver operating characteristic (ROC) curve, a widely used measure that evaluates a model’s ability to distinguish between classes (stillbirth vs. live birth). AUC values closer to 1 indicate better discrimination.
- Specificity and Sensitivity: Specificity (the true negative rate) measures how well the model identifies live births, while sensitivity (the true positive rate) assesses its ability to correctly identify stillbirths.
- Positive Predictive Value (PPV) and Negative Predictive Value (NPV): These metrics evaluate the proportion of true positive and true negative predictions, respectively, out of all positive and negative predictions made by the model.
2.4. Statistical Analysis
3. Results
3.1. Study Participants
3.2. Stillbirth Prediction
4. Discussion
4.1. The Study’s Main Outcomes
4.2. Comparative Analysis of Other Studies for Stillbirth Risk Factors
4.3. Stillbirth Prediction with Machine Learning in the Literature
4.4. Limitations
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Leduc, L. No. 394-Stillbirth Investigation. J. Obstet. Gynaecol. Can. 2020, 42, 92–99. [Google Scholar] [CrossRef] [PubMed]
- Boo, Y.Y.; Bora, A.K.; Chhabra, S.; Choudhury, S.S.; Deka, G.; Kakoty, S.; Kumar, P.; Mahanta, P.; Minz, B.; Rani, A.; et al. Maternal and fetal factors associated with stillbirth in singleton pregnancies in 13 hospitals across six states in India: A prospective cohort study. Int. J. Gynaecol. Obstet. Off. Organ Int. Fed. Gynaecol. Obstet. 2024, 165, 462–473. [Google Scholar] [CrossRef] [PubMed]
- Fretts, R. Stillbirth epidemiology, risk factors, and opportunities for stillbirth prevention. Clin. Obstet. Gynecol. 2010, 53, 588–596. [Google Scholar] [CrossRef] [PubMed]
- Heazell, A.E.P.; Siassakos, D.; Blencowe, H.; Burden, C.; Bhutta, Z.A.; Cacciatore, J.; Dang, N.; Das, J.; Flenady, V.; Gold, K.J.; et al. Stillbirths: Economic and psychosocial consequences. Lancet 2016, 387, 604–616. [Google Scholar] [CrossRef]
- Townsend, R.; Sileo, F.G.; Allotey, J.; Dodds, J.; Heazell, A.; Jorgensen, L.; Kim, V.B.; Magee, L.; Mol, B.; Sandall, J.; et al. Prediction of stillbirth: An umbrella review of evaluation of prognostic variables. BJOG Int. J. Obstet. Gynaecol. 2021, 128, 238–250. [Google Scholar] [CrossRef]
- Gardosi, J.; Madurasinghe, V.; Williams, M.; Malik, A.; Francis, A. Maternal and fetal risk factors for stillbirth: Population based study. BMJ (Clin. Res. Ed.) 2013, 346, f108. [Google Scholar] [CrossRef]
- Gibbins, K.J.; Pinar, H.; Reddy, U.M.; Saade, G.R.; Goldenberg, R.L.; Dudley, D.J.; Drews-Botsch, C.; Freedman, A.A.; Daniels, L.M.; Parker, C.B.; et al. Findings in Stillbirths Associated with Placental Disease. Am. J. Perinatol. 2020, 37, 708–715. [Google Scholar] [CrossRef]
- Goldenberg, R.L.; Kirby, R.; Culhane, J.F. Stillbirth: A review. J. Matern. Fetal Neonatal Med. 2004, 16, 79–94. [Google Scholar] [CrossRef]
- Darcy, A.M.; Louie, A.K.; Roberts, L.W. Machine Learning and the Profession of Medicine. JAMA 2016, 315, 551–552. [Google Scholar] [CrossRef]
- Smith, G.C. Predicting antepartum stillbirth. Curr. Opin. Obstet. Gynecol. 2006, 18, 625–630. [Google Scholar] [CrossRef]
- Pauli, R.M. Stillbirth: Fetal disorders. Clin. Obstet. Gynecol. 2010, 53, 646–655. [Google Scholar] [CrossRef]
- Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
- Cortes, C.; Vapnik, V. Support-vector network. Mach. Learn. 1995, 20, 273–297. [Google Scholar] [CrossRef]
- Yerlikaya, G.; Akolekar, R.; McPherson, K.; Syngelaki, A.; Nicolaides, K.H. Prediction of stillbirth from maternal demographic and pregnancy characteristics. Ultrasound Obstet. Gynecol. Off. J. Int. Soc. Ultrasound Obstet. Gynecol. 2016, 48, 607–612. [Google Scholar] [CrossRef]
- Yao, R.; Ananth, C.V.; Park, B.Y.; Pereira, L.; Plante, L.A. Obesity and the risk of stillbirth: A population-based cohort study. Am. J. Obstet. Gynecol. 2014, 210, 457.e1–457.e9. [Google Scholar] [CrossRef] [PubMed]
- Aune, D.; Saugstad, O.D.; Henriksen, T.; Tonstad, S. Maternal body mass index and the risk of fetal death, stillbirth, and infant death: A systematic review and meta-analysis. JAMA 2014, 311, 1536–1546. [Google Scholar] [CrossRef]
- Marufu, T.C.; Ahankari, A.; Coleman, T.; Lewis, S. Maternal smoking and the risk of still birth: Systematic review and meta-analysis. BMC Public Health 2015, 15, 239. [Google Scholar] [CrossRef]
- Hansen, A.T.; Schmidt, M.; Horváth-Puhó, E.; Pedersen, L.; Rothman, K.J.; Hvas, A.M.; Sørensen, H.T. Preconception venous thromboembolism and placenta-mediated pregnancy complications. J. Thromb. Haemost. JTH 2015, 13, 1635–1641. [Google Scholar] [CrossRef]
- Tsakiridis, I.; Giouleka, S.; Mamopoulos, A.; Athanasiadis, A.; Dagklis, T. Investigation and management of stillbirth: A descriptive review of major guidelines. J. Perinat. Med. 2022, 50, 796–813. [Google Scholar] [CrossRef]
- Nijkamp, J.W.; Korteweg, F.J.; Groen, H.; Timmer, A.; Van Den Berg, G.; Bossuyt, P.M.; Mol, B.W.; Erwich, J.J. Thyroid function testing in women who had a stillbirth. Clin. Endocrinol. 2016, 85, 291–298. [Google Scholar] [CrossRef]
- Hayes, D.J.L.; Warland, J.; Parast, M.M.; Bendon, R.W.; Hasegawa, J.; Banks, J.; Clapham, L.; Heazell, A.E.P. Umbilical cord characteristics and their association with adverse pregnancy outcomes: A systematic review and meta-analysis. PLoS ONE 2020, 15, e0239630. [Google Scholar] [CrossRef] [PubMed]
- Deng, X.; Pan, B.; Lai, H.; Sun, Q.; Lin, X.; Yang, J.; Han, X.; Ge, T.; Li, Q.; Ge, L.; et al. Association of previous stillbirth with subsequent perinatal outcomes: A systematic review and meta-analysis of cohort studies. Am. J. Obstet. Gynecol. 2024, 231, 211–222. [Google Scholar] [CrossRef] [PubMed]
- Frey, H.A.; Odibo, A.O.; Dicke, J.M.; Shanks, A.L.; Macones, G.A.; Cahill, A.G. Stillbirth risk among fetuses with ultrasound-detected isolated congenital anomalies. Obstet. Gynecol. 2014, 124, 91–98. [Google Scholar] [CrossRef]
- Hoyert, D.L.; Gregory, E.C.W. Cause-of-death Data From the Fetal Death File, 2015–2017. Natl. Vital Stat. Rep. 2020, 69, 1–20. [Google Scholar] [PubMed]
- The Stillbirth Collaborative Research Network Writing Group. Causes of death among stillbirths. Jama 2011, 306, 2459–2468. [Google Scholar] [CrossRef]
- Rabie, N.; Magann, E.; Steelman, S.; Ounpraseuth, S. Oligohydramnios in complicated and uncomplicated pregnancy: A systematic review and meta-analysis. Ultrasound Obstet. Gynecol. 2017, 49, 442–449. [Google Scholar] [CrossRef]
- Casey, B.M.; McIntire, D.D.; Bloom, S.L.; Lucas, M.J.; Santos, R.; Twickler, D.M.; Ramus, R.M.; Leveno, K.J. Pregnancy outcomes after antepartum diagnosis of oligohydramnios at or beyond 34 weeks’ gestation. Am. J. Obstet. Gynecol. 2000, 182, 909–912. [Google Scholar] [CrossRef]
- Hadar, E.; Melamed, N.; Sharon-Weiner, M.; Hazan, S.; Rabinerson, D.; Glezerman, M.; Yogev, Y. The association between stillbirth and fetal gender. J. Matern. Fetal Neonatal Med. 2012, 25, 158–161. [Google Scholar] [CrossRef]
- Maghsoudlou, S.; Cnattingius, S.; Aarabi, M.; Montgomery, S.M.; Semnani, S.; Stephansson, O.; Wikström, A.K.; Bahmanyar, S. Consanguineous marriage, prepregnancy maternal characteristics and stillbirth risk: A population-based case-control study. Acta Obstet. Gynecol. Scand. 2015, 94, 1095–1101. [Google Scholar] [CrossRef]
- Nybo Andersen, A.M.; Gundlund, A.; Villadsen, S.F. Stillbirth and congenital anomalies in migrants in Europe. Best Pract. Res. Clin. Obstet. Gynaecol. 2016, 32, 50–59. [Google Scholar] [CrossRef]
- Kapurubandara, S.; Melov, S.; Shalou, E.; Alahakoon, I. Consanguinity and associated perinatal outcomes, including stillbirth. Aust. N. Z. J. Obstet. Gynaecol. 2016, 56, 599–604. [Google Scholar] [CrossRef] [PubMed]
- Lawn, J.E.; Blencowe, H.; Waiswa, P.; Amouzou, A.; Mathers, C.; Hogan, D.; Flenady, V.; Frøen, J.F.; Qureshi, Z.U.; Calderwood, C.; et al. Stillbirths: Rates, risk factors, and acceleration towards 2030. Lancet 2016, 387, 587–603. [Google Scholar] [CrossRef] [PubMed]
- Vogel, J.P.; Souza, J.P.; Mori, R.; Morisaki, N.; Lumbiganon, P.; Laopaiboon, M.; Ortiz-Panozo, E.; Hernandez, B.; Pérez-Cuevas, R.; Roy, M.; et al. Maternal complications and perinatal mortality: Findings of the World Health Organization Multicountry Survey on Maternal and Newborn Health. BJOG 2014, 121 (Suppl. S1), 76–88. [Google Scholar] [CrossRef]
- Xiong, T.; Mu, Y.; Liang, J.; Zhu, J.; Li, X.; Li, J.; Liu, Z.; Qu, Y.; Wang, Y.; Mu, D. Hypertensive disorders in pregnancy and stillbirth rates: A facility-based study in China. Bull. World Health Organ. 2018, 96, 531–539. [Google Scholar] [CrossRef]
- Li, Q.; Li, P.; Chen, J.; Ren, R.; Ren, N.; Xia, Y. Machine Learning for Predicting Stillbirth: A Systematic Review. Reprod. Sci. 2024, online. [Google Scholar] [CrossRef]
- Cersonsky, T.E.K.; Ayala, N.K.; Pinar, H.; Dudley, D.J.; Saade, G.R.; Silver, R.M.; Lewkowitz, A.K. Identifying risk of stillbirth using machine learning. Am. J. Obstet. Gynecol. 2023, 229, 327.e1–327.e16. [Google Scholar] [CrossRef]
- Khatibi, T.; Hanifi, E.; Sepehri, M.M.; Allahqoli, L. Proposing a machine-learning based method to predict stillbirth before and during delivery and ranking the features: Nationwide retrospective cross-sectional study. BMC Pregnancy Childbirth 2021, 21, 202. [Google Scholar] [CrossRef]
- Malacova, E.; Tippaya, S.; Bailey, H.D.; Chai, K.; Farrant, B.M.; Gebremedhin, A.T.; Leonard, H.; Marinovich, M.L.; Nassar, N.; Phatak, A.; et al. Stillbirth risk prediction using machine learning for a large cohort of births from Western Australia, 1980–2015. Sci. Rep. 2020, 10, 5354. [Google Scholar] [CrossRef]
Parameters | Stillbirth Group (n:452) | Live Birth Group (n:499) | p Value | |
---|---|---|---|---|
Maternal age | 30.22 ± 6.33 | 30.24 ± 5.84 | 0.961 * | |
BMI (body mass index) | 28.84 ± 3.28 | 27.70 ± 4.62 | 0.001 * | |
Gravida | 3 (1–10) | 3 (1–11) | 0.802 ** | |
Parity | 1 (0–7) | 1 (0–6) | 0.647 ** | |
Previous miscarriage | 0 (0–7) | 0 (0–7) | 0.633 ** | |
Previous cesarean section count | 0.78 ± 0.98 | 0.54 ± 0.90 | 0.001 * | |
Previous vaginal delivery count | 0.56 ± 0.99 | 0.79 ± 1.13 | 0.001 * | |
Induction methods used | Oxytocin | 86 (19%) | 34 (6.8%) | 0.001 *** |
Misoprostol | 164 (36.3%) | 34 (6.8%) | 0.001 *** | |
Oxytocin + misoprostol | 24 (5.3%) | 34 (6.8%) | 0.333 *** | |
Birth week | 27.05 ± 5.51 | 37.88 ± 1.86 | 0.001 * | |
Delivery type | VD | 303 (67%) | 100 (20%) | 0.001 *** |
CS | 149 (33%) | 399 (80%) | ||
Gender | Girl | 187 (41.4%) | 237 (47.5%) | 0.223 |
Boy | 242 (53.5%) | 261 (52.3%) | ||
Ambiguous genitalia | 23 (5.1%) | 1 (0.2%) | 0.001 *** |
Parameters | Stillbirth Group (n:452) | Live Birth Group (n:499) | p Value | ||
---|---|---|---|---|---|
Consanguinity | Present | 44 (9.7%) | 10 (2%) | 0.001 *** | |
Fetal anomalies | Present | 150 (33.2%) | 34 (6.8%) | 0.001 *** | |
Primiparity | Present | 153 (33.8%) | 146 (29.3%) | 0.128 | |
IVF | Present | 10 (2.2%) | 14 (2.8%) | 0.560 *** | |
RH/rh | Present | 31 (6.9%) | 49 (9.8%) | 0.100 *** | |
Stillbirth history | Present | 35/7.7%) | 7 (1.4%) | 0.001 *** | |
Cigarette | Present | 4 (0.9%) | 0 (0%) | 0.035 *** | |
PPROM | Present | 62 (13.7%) | 52 (10.4%) | 0.118 *** | |
Preeclampsia | Present | 37 (8.2%) | 22 (4.4%) | 0.016 *** | |
Chronic hypertension | Present | 12 (2.7%) | 4 (0.8%) | 0.026 *** | |
Pre-existing diabetes mellitus | Present | 16 (3.5%) | 27 (5.4%) | 0.166 *** | |
GDM | Present | 11 (2.4%) | 20 (4%) | 0.172 *** | |
Cholestasis | Present | 1 (0.2%) | 1 (0.2%) | 0.944 *** | |
History of thrombosis | Present | 49 (10.8%) | 6 (1.2%) | 0.001 *** | |
Maternal heart disease | Present | 17 (3.8%) | 6 (1.2%) | 0.010 *** | |
APS | Present | 1 (0.2%) | 3 (0.6%) | 0.366 *** | |
Placenta previa | Present | 18 (4%) | 12 (2.4%) | 0.165 *** | |
Ablatio placenta | Present | 16 (3.5%) | 2 (0.4%) | 0.001 *** | |
Placenta accreta spectrum | Present | 5 (1.1%) | 5 (1.0%) | 0.875 *** | |
Polyhydramnios | Present | 17 (3.8%) | 17 (3.4%) | 0.769 *** | |
Oligohydramnios | Present | 103 (22.8%) | 63 (12.6%) | 0.001 *** | |
Fetal growth restriction | Present | 62 (13.7%) | 51 (10.2%) | 0.096 *** | |
Cord prolapse | Present | 1 (0.2%) | 1 (0.2%) | 0.944 *** | |
Cord knot | Present | 1 (0.2%) | 0 (0%) | 0.293 *** | |
Nuchal cord | Present | 11 (2.4%) | 8 (1.6%) | 0.361 *** | |
Thyroid disease | Hypothyroidism | Present | 59 (13.3%) | 71 (14.3%) | 0.678 *** |
Hyperthyroidism | Present | 10 (2.5%) | 2 (0.5%) | 0.013 *** |
Univariable | Multivariable | |||
---|---|---|---|---|
Variable | OR (95%CI) | p | OR (95%CI) | p |
BMI | 1.07 (1.03–1.11) | 0.001 * | 1.07 (1.03–1.12) | 0.001 * |
Consanguinity | 5.27 (2.62–10.61) | 0.001 *** | 6.14 (2.80–13.43) | 0.001 *** |
History of stillbirth | 5.89 (2.59–13.42) | 0.001 *** | 7.31 (2.76–19.31) | 0.001 *** |
Fetal anomalies | 6.79 (4.55–10.12) | 0.001 *** | 8.50 (5.54–13.03) | 0.001 *** |
EMR | 1.36 (0.92–2.02) | 0.119 | ||
Preeclampsia | 1.93 (1.12–3.33) | 0.018 *** | 2.12 (1.13–3.95) | 0.024 *** |
Chronic hypertension | 3.37 (1.08–10.54) | 0.036 *** | 3.25 (0.82–12.85) | 0.092 *** |
Diabetes mellitus | 1.55 (0.829–2.93) | 0.169 | ||
GDM | 1.67 (0.79–3.53) | 0.176 | ||
Cholestasis | 1.10 (0.06–17.70) | 0.944 | ||
History of thrombosis | 10.84 (4.29–27.37) | 0.001 *** | 14.13 (5.08–39.31) | 0.001 *** |
Maternal heart disease | 3.53 (1.31–9.55) | 0.013 *** | 3.79 (1.25–11.42) | 0.018 *** |
Hyperthyroidism | 5.56 (1.21–25.54) | 0.027 | 1.13 (0.73–1,75) | 0.576 |
APS | 2.72 (0.28–26.31) | 0.386 | ||
Placental abruption | 9.11 (2.08–39.88) | 0.003 *** | 12.76 (2.28–71.22) | 0.004 *** |
Placenta percreta | 1.15 (0.29–4.46) | 0.837 | ||
Polyhydramnios | 1.02 (0.47–2.24) | 0.945 | ||
Oligohydramnios | 2.56 (1.77–3.70) | 0.001 *** | 2.46 (1.65–3.65) | 0.001 *** |
Fetal growth restriction | 1.36 (0.89–2.09) | 0.153 | ||
Nuchal cord | 1.53 (0.61–3.84) | 0.364 |
Logistic Regression | Random Forest | Support Vector Machine | Multi-Layer Perceptron | |
---|---|---|---|---|
Accuracy | 94.24% | 96.86% | 95.29% | 94.24% |
Sensitivity | 93.90% | 96.34% | 91.46% | 92.68% |
Specificity | 94.50% | 97.25% | 98.17% | 95.41% |
PPV | 92.77% | 96.34% | 97.40% | 93.83% |
NPV | 95.37% | 97.25% | 93.86% | 94.55% |
AUC (%95 CI) | 0.98 (0.96–1.00) | 0.99 (0.97–1.00) | 0.98 (0.96–1.00) | 0.97 (0.95–1.00) |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Published by MDPI on behalf of the Lithuanian University of Health Sciences. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Gunenc, O.; Dogru, S.; Yaman, F.K.; Ezveci, H.; Metin, U.S.; Acar, A. The Application of Machine Learning Models to Predict Stillbirths. Medicina 2025, 61, 472. https://doi.org/10.3390/medicina61030472
Gunenc O, Dogru S, Yaman FK, Ezveci H, Metin US, Acar A. The Application of Machine Learning Models to Predict Stillbirths. Medicina. 2025; 61(3):472. https://doi.org/10.3390/medicina61030472
Chicago/Turabian StyleGunenc, Oguzhan, Sukran Dogru, Fikriye Karanfil Yaman, Huriye Ezveci, Ulfet Sena Metin, and Ali Acar. 2025. "The Application of Machine Learning Models to Predict Stillbirths" Medicina 61, no. 3: 472. https://doi.org/10.3390/medicina61030472
APA StyleGunenc, O., Dogru, S., Yaman, F. K., Ezveci, H., Metin, U. S., & Acar, A. (2025). The Application of Machine Learning Models to Predict Stillbirths. Medicina, 61(3), 472. https://doi.org/10.3390/medicina61030472